Acoustic monitoring of compaction in cohesive granular materials.

We study the transition from cohesive to noncohesive states of cemented granular materials (synthetic rocks) under oedometric loading, combining simultaneous measurements of ultrasound velocity and acoustic emission (AE: microseosmicity). Our samples are agglomerates made of glass beads bonded with a few percent of cement, either ductile or brittle. These cemented granular samples exhibit an inelastic compaction beyond certain axial stresses likely due to the formation of compaction bands, which is accompanied by a significant decrease of compressional wave velocity. Upon subsequent cyclic unloading-reloading with constant consolidation stress, we found the mechanical and acoustic responses like those in noncohesive granular materials, which can be interpreted within the effective medium theory based on the Digby's bonding model. Moreover, this model allows P-wave velocity measured at vanishing pressure to be interpreted as an indicator of the debonding on the scale of grain contact. During the inelastic compaction, stick-slip-like stress drops were observed in brittle cement-bonded granular samples accompanied by the instantaneous decrease of the P-wave velocity and AEs which display an Omori-like law for foreshocks, i.e., precursors. By contrast, mechanical responses of ductile cement-bonded granular samples are smooth (without visible stick-slip-like stress drops) and mostly aseismic. By applying a cyclic loading-unloading with increasing consolidation stress, we observed a Kaiser-like memory effect in the brittle cement-bonded sample in the weakly damaged state which tends to disappear when the bonds are mostly broken in the noncohesive granular state after large-amplitude loading. In this paper, we show that the macroscopic ductile and brittle behavior of cemented granular media is controlled by the local processes on the scale of the bonds between grains.

AI-Based Unmixing of Medium and Source Signatures From Seismograms: Ground Freezing Patterns.

Seismograms always result from mixing many sources and medium changes that are complex to disentangle, witnessing many physical phenomena within the Earth. With artificial intelligence (AI), we isolate the signature of surface freezing and thawing in continuous seismograms recorded in a noisy urban environment. We perform a hierarchical clustering of the seismograms and identify a pattern that correlates with ground frost periods. We further investigate the fingerprint of this pattern and use it to track the continuous medium change with high accuracy and resolution in time. Our method isolates the effect of the ground frost and describes how it affects the horizontal wavefield. Our findings show how AI-based strategies can help to identify and understand hidden patterns within seismic data caused either by medium or source changes.

Space-time monitoring of groundwater fluctuations with passive seismic interferometry.

Historic levels of drought, globally, call for sustainable freshwater management. Under pressing demand is a refined understanding of the structures and dynamics of groundwater systems. Here we present an unconventional, cost-effective approach to aquifer monitoring using seismograph arrays. Employing advanced seismic interferometry techniques, we calculate the space-time evolution of relative changes in seismic velocity, as a measure of hydrological properties. During 2000-2020 in basins near Los Angeles, seismic velocity variations match groundwater tables measured in wells and surface deformations inferred from satellite sensing, but the seismological approach adds temporal and depth resolutions for deep structures and processes. Maps of long-term seismic velocity changes reveal distinct patterns (decline or recovery) of groundwater storage in basins that are adjacent but adjudicated to water districts conducting different pumping practices. This pilot application bridges the gap between seismology and hydrology, and shows the promise of leveraging seismometers worldwide to provide 4D characterizations of groundwater and other near-surface systems.

Hierarchical Exploration of Continuous Seismograms With Unsupervised Learning.

Continuous seismograms contain a wealth of information with a large variety of signals with different origin. Identifying these signals is a crucial step in understanding physical geological objects. We propose a strategy to identify classes of signals in continuous single-station seismograms in an unsupervised fashion. Our strategy relies on extracting meaningful waveform features based on a deep scattering network combined with an independent component analysis. Based on the extracted features, agglomerative clustering then groups these waveforms in a hierarchical fashion and reveals the process of clustering in a dendrogram. We use the dendrogram to explore the seismic data and identify different classes of signals. To test our strategy, we investigate a two-day-long seismogram collected in the vicinity of the North Anatolian Fault, Turkey. We analyze the automatically inferred clusters' occurrence rate, spectral characteristics, cluster size, and waveform and envelope characteristics. At a low level in the cluster hierarchy, we obtain three clusters related to anthropogenic and ambient seismic noise and one cluster related to earthquake activity. At a high level in the cluster hierarchy, we identify a seismic burst that includes around 200 events with similar waveforms and high-frequent signals with correlating envelopes and an anthropogenic origin. The application shows that the cluster hierarchy helps to identify particular families of signals and to extract subclusters for further analysis. This is valuable when certain types of signals, such as earthquakes, are under-represented in the data. The proposed method may also successfully discover new types of signals since it is entirely data-driven.

Clustering earthquake signals and background noises in continuous seismic data with unsupervised deep learning.

The continuously growing amount of seismic data collected worldwide is outpacing our abilities for analysis, since to date, such datasets have been analyzed in a human-expert-intensive, supervised fashion. Moreover, analyses that are conducted can be strongly biased by the standard models employed by seismologists. In response to both of these challenges, we develop a new unsupervised machine learning framework for detecting and clustering seismic signals in continuous seismic records. Our approach combines a deep scattering network and a Gaussian mixture model to cluster seismic signal segments and detect novel structures. To illustrate the power of the framework, we analyze seismic data acquired during the June 2017 Nuugaatsiaq, Greenland landslide. We demonstrate the blind detection and recovery of the repeating precursory seismicity that was recorded before the main landslide rupture, which suggests that our approach could lead to more informative forecasting of the seismic activity in seismogenic areas.

Slow slip events in the roots of the San Andreas fault.

Episodic tremor and accompanying slow slip are observed at the down-dip edge of subduction seismogenic zones. While tremors are the seismic signature of this phenomenon, they correspond to a small fraction of the moment released; thus, the associated fault slip can be quantified only by geodetic observations. On continental strike-slip faults, tremors have been observed in the roots of the Parkfield segment of the San Andreas fault. However, associated transient aseismic slip has never been detected. By making use of the timing of transient tremor activity and the dense Parkfield-area global positioning system network, we can detect deep slow slip events (SSEs) at 16-km depth on the Parkfield segment with an average moment equivalent to M w 4.90 ± 0.08. Characterization of transient SSEs below the Parkfield locked asperity, at the transition with the creeping section of the San Andreas fault, provides new constraints on the seismic cycle in this region.

Revealing the cluster of slow transients behind a large slow slip event.

Capable of reaching similar magnitudes to large megathrust earthquakes [Mw (moment magnitude) > 7], slow slip events play a major role in accommodating tectonic motion on plate boundaries through predominantly aseismic rupture. We demonstrate here that large slow slip events are a cluster of short-duration slow transients. Using a dense catalog of low-frequency earthquakes as a guide, we investigate the Mw 7.5 slow slip event that occurred in 2006 along the subduction interface 40 km beneath Guerrero, Mexico. We show that while the long-period surface displacement, as recorded by Global Positioning System, suggests a 6-month duration, the motion in the direction of tectonic release only sporadically occurs over 55 days, and its surface signature is attenuated by rapid relocking of the plate interface. Our proposed description of slow slip as a cluster of slow transients forces us to re-evaluate our understanding of the physics and scaling of slow earthquakes.

Compositional heterogeneity near the base of the mantle transition zone beneath Hawaii.

Global seismic discontinuities near 410 and 660 km depth in Earth's mantle are expressions of solid-state phase transitions. These transitions modulate thermal and material fluxes across the mantle and variations in their depth are often attributed to temperature anomalies. Here we use novel seismic array analysis of SS waves reflecting off the 410 and 660 below the Hawaiian hotspot. We find amplitude-distance trends in reflectivity that imply lateral variations in wavespeed and density contrasts across 660 for which thermodynamic modeling precludes a thermal origin. No such variations are found along the 410. The inferred 660 contrasts can be explained by mantle composition varying from average (pyrolitic) mantle beneath Hawaii to a mixture with more melt-depleted harzburgite southeast of the hotspot. Such compositional segregation was predicted, from petrological and numerical convection studies, to occur near hot deep mantle upwellings like the one often invoked to cause volcanic activity on Hawaii.

The evolving interaction of low-frequency earthquakes during transient slip.

Observed along the roots of seismogenic faults where the locked interface transitions to a stably sliding one, low-frequency earthquakes (LFEs) primarily occur as event bursts during slow slip. Using an event catalog from Guerrero, Mexico, we employ a statistical analysis to consider the sequence of LFEs at a single asperity as a point process, and deduce the level of time clustering from the shape of its autocorrelation function. We show that while the plate interface remains locked, LFEs behave as a simple Poisson process, whereas they become strongly clustered in time during even the smallest slow slip, consistent with interaction between different LFE sources. Our results demonstrate that bursts of LFEs can result from the collective behavior of asperities whose interaction depends on the state of the fault interface.

Green's function retrieval through cross-correlations in a two-dimensional complex reverberating medium.

Cross-correlations of ambient noise averaged at two receivers lead to the reconstruction of the two-point Green's function, provided that the wave-field is uniform azimuthally, and also temporally and spatially uncorrelated. This condition depends on the spatial distribution of the sources and the presence of heterogeneities that act as uncorrelated secondary sources. This study aims to evaluate the relative contributions of source distribution and medium complexity in the two-point cross-correlations by means of numerical simulations and laboratory experiments in a finite-size reverberant two-dimensional (2D) plate. The experiments show that the fit between the cross-correlation and the 2D Green's function depends strongly on the nature of the source used to excite the plate. A turbulent air-jet produces a spatially uncorrelated acoustic field that rapidly builds up the Green's function. On the other hand, extracting the Green's function from cross-correlations of point-like sources requires more realizations and long recordings to balance the effect of the most energetic first arrivals. When the Green's function involves other arrivals than the direct wave, numerical simulations confirm the better Green's function reconstruction with a spatially uniform source distribution than the typical contour-like source distribution surrounding the receivers that systematically gives rise to spurious phases.

A passive inverse filter for Green's function retrieval.

Passive methods for the recovery of Green's functions from ambient noise require strong hypotheses, including isotropic distribution of the noise sources. Very often, this distribution is nonisotropic, which introduces bias in the Green's function reconstruction. To minimize this bias, a spatiotemporal inverse filter is proposed. The method is tested on a directive noise field computed from an experimental active seismic data set. The results indicate that the passive inverse filter allows the manipulation of the spatiotemporal degrees of freedom of a complex wave field, and it can efficiently compensate for the noise wavefield directivity.

On the correlation of non-isotropically distributed ballistic scalar diffuse waves.

Theorems indicating that a fully equipartitioned random wave field will have correlations equivalent to the Green's function that would be obtained in an active measurement are now legion. Studies with seismic waves, ocean acoustics, and laboratory ultrasound have confirmed them. So motivated, seismologists have evaluated apparent seismic travel times in correlations of ambient seismic noise and tomographically constructed impressive maps of seismic wave velocity. Inasmuch as the random seismic waves used in these evaluations are usually not fully equipartitioned, it seems right to ask why it works so well, or even if the results are trustworthy. The error, in apparent travel time, due to non-isotropic specific intensity is evaluated here in a limit of large receiver-receiver separation and for the case in which the source of the noise is in the far field of both receivers. It is shown that the effect is small, even for cases in which one might have considered the anisotropy to be significant, and even for station pairs separated by as little as one or two wavelengths. A formula is derived that permits estimations of error and corrections to apparent travel time. It is successfully compared to errors seen in synthetic waveforms.

Stability of monitoring weak changes in multiply scattering media with ambient noise correlation: laboratory experiments.

Previous studies have shown that small changes can be monitored in a scattering medium by observing phase shifts in the coda. Passive monitoring of weak changes through ambient noise correlation has already been applied to seismology, acoustics, and engineering. Usually, this is done under the assumption that a properly reconstructed Green function (GF), as well as stable background noise sources, is necessary. In order to further develop this monitoring technique, a laboratory experiment was performed in the 2.5 MHz range in a gel with scattering inclusions, comparing an active (pulse-echo) form of monitoring to a passive (correlation) one. Present results show that temperature changes in the medium can be observed even if the GF of the medium is not reconstructed. Moreover, this article establishes that the GF reconstruction in the correlations is not a necessary condition: The only condition to monitoring with correlation (passive experiment) is the relative stability of the background noise structure.

Small-scale seismic inversion using surface waves extracted from noise cross correlation.

Green's functions can be retrieved between receivers from the correlation of ambient seismic noise or with an appropriate set of randomly distributed sources. This principle is demonstrated in small-scale geophysics using noise sources generated by human steps during a 10-min walk in the alignment of a 14-m-long accelerometer line array. The time-domain correlation of the records yields two surface wave modes extracted from the Green's function between each pair of accelerometers. A frequency-wave-number Fourier analysis yields each mode contribution and their dispersion curve. These dispersion curves are then inverted to provide the one-dimensional shear velocity of the near surface.

Reconstruction of Rayleigh-Lamb dispersion spectrum based on noise obtained from an air-jet forcing.

The time-domain cross correlation of incoherent and random noise recorded by a series of passive sensors contains the impulse response of the medium between these sensors. By using noise generated by a can of compressed air sprayed on the surface of a plexiglass plate, we are able to reconstruct not only the time of flight but the whole wave forms between the sensors. From the reconstruction of the direct A(0) and S(0) waves, we derive the dispersion curves of the flexural waves, thus estimating the mechanical properties of the material without a conventional electromechanical source. The dense array of receivers employed here allow a precise frequency-wavenumber study of flexural waves, along with a thorough evaluation of the rate of convergence of the correlation with respect to the record length, the frequency, and the distance between the receivers. The reconstruction of the actual amplitude and attenuation of the impulse response is also addressed in this paper.

Observation of multiple scattering of kHz vibrations in a concrete structure and application to monitoring weak changes.

In this paper, we present two experimental studies of mechanical wave propagation in a concrete building around 1 kHz. The first experiment is devoted to the observation of the coherent backscattering enhancement, which demonstrates the presence of multiple diffractions in the late part of the wave records. An application of multiple diffraction and reverberations is proposed in a second experiment. Thanks to their sensitivity to weak changes of the medium, the late records are used to monitor weak change in concrete wave velocity induced by thermal variations. The velocity change measurements have a precision of deltac/c=10(-4). Such a precision is difficult to obtain with direct waves. This experiment is the first step to other applications like stress, damage, aging, or crack monitoring in concrete structures.

Passive retrieval of Rayleigh waves in disordered elastic media.

When averaged over sources or disorder, cross correlation of diffuse fields yields the Green's function between two passive sensors. This technique is applied to elastic ultrasonic waves in an open scattering slab mimicking seismic waves in the Earth's crust. It appears that the Rayleigh wave reconstruction depends on the scattering properties of the elastic slab. Special attention is paid to the specific role of bulk to Rayleigh wave coupling, which may result in unexpected phenomena, such as a persistent time asymmetry in the diffuse regime.

High-resolution surface-wave tomography from ambient seismic noise.

Cross-correlation of 1 month of ambient seismic noise recorded at USArray stations in California yields hundreds of short-period surface-wave group-speed measurements on interstation paths. We used these measurements to construct tomographic images of the principal geological units of California, with low-speed anomalies corresponding to the main sedimentary basins and high-speed anomalies corresponding to the igneous cores of the major mountain ranges. This method can improve the resolution and fidelity of crustal images obtained from surface-wave analyses.

Long-range correlations in the diffuse seismic coda.

The late seismic coda may contain coherent information about the elastic response of Earth. We computed the correlations of the seismic codas of 101 distant earthquakes recorded at stations that were tens of kilometers apart. By stacking cross-correlation functions of codas, we found a low-frequency coherent part in the diffuse field. The extracted pulses have the polarization characteristics and group velocities expected for Rayleigh and Love waves. The set of cross-correlations has the symmetries of the surface-wave part of the Green tensor. This seismological example shows that diffuse waves produced by distant sources are sufficient to retrieve direct waves between two perfectly located points of observation. Because it relies on general properties of diffuse waves, this result has potential applications in other fields.